In serial femtosecond crystallography (SFX) with hard X-ray free-electron laser as light source, a set of three-dimensional single-crystal
diffraction data is produced by a huge number of crystal grains in micron size --- a typical powder sample. This is beneficial to protein
crystallography by avoiding growth of large single crystal samples. This may also capable of enhancing the power of powder method, in solving
crystal structures, from ~20 independent atoms to more than a thousand. Another important feature of SFX experiments is that diffraction
patterns of a single experiment for multi-phase samples can be identified and grouped to independent subsets corresponding to each single
phase. Simulating calculations were performed to exploring possibilities of SFX experiments. Encouraging results have been obtained.
Phase identification and crystal-structure determination from a
single SFX experiment for a mixture sample of two zeolites
Sample: 4(TNU-9) + NU-87
TNU-9:
Unit cell contents: Na, 24(SiO2); space group: C2/m;
Cell dimensions: a=27.845 Å, b=20.015 Å, c=19.597 Å, β=93.000°
NU-87:
Unit cell contents: 17(SiO2); space group: P21/c;
Cell dimensions: a=14.324 Å, b=22.376 Å, c=25.092 Å, β=151.51°
These are very closed respectively to the known values of TNU-9 and
NU-87. Crystal structures of the two components were determined
using the program SHELXD. Projections of the two structures are
shown in the following figure. As can be seen, results from SHELXD
are nearly complete, except the type of a few atoms should be
corrected.
Simulation: The program pattern_sim in crystFEL package was used to
simulate SFX diffraction data. Wavelength of X-ray used in the
simulation is 0.7 Å; X-ray intensity is 1011 photons per pulse; size of
crystals is from 600 nm to 2000 nm. Two samples were calculated
independently firstly. Then the simulation patterns were mixed
randomly with a ratio of 4 to 1. One million of diffraction patterns were
calculated in total, and 799139 of which belong to TNU-9.
SAD: phasing with impure sample
Identification: Single diffraction patterns were indexed separately by
the program indexamajig in CrystFEL without information about cell
dimensions. Each pattern that can be identified successfully defined a
set of cell dimensions belong to itself. Volume of unit cell was chosen
as an indicator to represent each set of cell dimensions. The distribution
of number of patterns vs. unit cell volume was drawn as shown below.
The two major peaks at 3800Å3 and 5500Å3 can be identified easily.
Small peaks represent larger volume that are multiple of either of the
two major peaks. Blue peaks belong to component 1; red ones belong to
component 2. So the two kinds of components in the sample can be
clearly identified.
In preparation of protein samples, especially heavy-atom derivatives for
SFX, it is unavoidable that native and derivative(s) are mixed together.
It is essential to extract diffraction patterns that belong to a single phase
from a vast number of diffraction patterns yielded by a mixture sample.
This can be done by a similar identification method proposed above.
Simulating calculations were performed with a set of SFX SAD data
from the mixture of Gadolinium derivative (90%) and native (10%) of
lysozyme. When the whole set of mixed data were used in a SAD
phasing, no reasonable result was obtained. However, if the subset of
derivative data extracted by the identification process were used, the
result is nearly a complete structure model.
SIR: native & derivative data from a single SFX experiment
After analyzing the results of indexing, cell dimensions of the two
components are refined as
(1) a=17.30 Å, b=17.34 Å, c=19.46 Å, α=87.89o, β=87.89 o, γ=70.84 o
(2) a=13.66 Å, b=14.25 Å, c=22.35 Å, α=89.90o, β=89.92 o, γ=61.87°
All the diffraction patterns were re-indexed by the program
indexamajig with information of cell dimensions obtained. The number
of patterns that belong to component (1) is 675185, while that of
component (2) is 159562. The ration is 4.40 : 1.
In the single isomorphous replacement (SIR) case, we can used a
mixture sample containing both native and derivative crystals. After the
SFX experiment, diffraction patterns can be identified and grouped into
two subsets, one corresponding to the native, while the other to the
derivative. We thus obtained a set of SIR data from a single SFX
experiment. Simulating calculation with the known protein LegC3N led
to a satisfactory result. Similar process may also be applied in the
multiple isomorphous replacement (MIR) case.
Structure determination: The program XPREP was used to rearrange
the cell dimensions and determine the space group. Results are as
bellow
(1) C2/m; a=28.23 Å, b=20.08 Å, c=19.46 Å,β=92.59o
(2) P21/c; a=14.32 Å, b=22.38 Å, c=25.09 Å,β=151.50o
E-mail: fanhf@cryst.iphy.ac.cn